SIAP: an intelligent algorithm for multiple prescription pattern recognition based on weighted similarity distances.

Area of Science:

• Computational medicine
• Pharmacology
• Medical informatics

Background:

• Complex diseases often involve multiple conditions and treatments, complicating clinical data analysis.
• Extracting knowledge from clinical data is challenging due to the multidimensionality of treatments.
• Traditional Chinese Medicine (TCM) offers a rich source of complex prescription data.

Purpose of the Study:

• To develop a novel algorithm for identifying subgroups within complex prescriptions.
• To determine the importance level of individual drugs in complex prescriptions.
• To automate the matching of complex drug sub-prescriptions with standard or classic prescriptions.

Main Methods:

• Proposed the Subgroup Identification Algorithm for Complex Prescriptions (SIAP).
• Applied SIAP to classify drug importance and identify valid prescription combinations.
• Validated the algorithm using classical TCM prescriptions and clinical herbal prescriptions.
• Optimized SIAP and its variants (SIAP-All, SIAP + All) using training and test sets.
• Compared performance against the baseline Intersection Set Rate (ISR) algorithm.

Main Results:

• SIAP-All and SIAP + All algorithms demonstrated superior performance over the ISR algorithm.
• Achieved improved accuracy, recall, and F1 values.
• SIAP + All yielded an F1 value of 0.7799, an 11.04% improvement over ISR.

Conclusions:

• Developed SIAP for automated matching of complex drug sub-prescriptions to standard ones.
• The algorithm effectively weights drugs based on their importance.
• Results facilitate classification and analysis of complex prescription drug compositions.